Research On Image Denoising And Edge Detection

With the development of digital image processing techniques, it's applied more and more extensively. Due to the influence of factors by noise in the process of acquisition and transmission, it makes the degradation of the quality of images. The noise characteristics of image and its recovery is one of the hot issues of the image-processing technology, the study of image denoising can help to the subsequent work of image edge extraction, and it's also the basic of the subsequent image processing. The image edge contains a lot of information; it has significant effect on image segmentation and recognition by using the edge detection effectively and accurately. Therefore, it has the vital significance to research on the image denoising and edge detection.This paper analyzed the current image denoising and edge detection processing technology, combined the algorithm of image denoising and edge detection in the aspect of airspace, and then presented some algorithms, through the theoretical analysis and experimental results, it proved that such algorithms are effective and feasible. The main results in this paper are as follows:(1) The scanning image denoising based on the Otsu methodScanning is one of the important ways to obtain image, while the image through scanning often remain shadows, it seriously affected the quality of the scanning images. Calculate the histogram using in the scanning images and the appeared probability of all the grayscale. According to the histogram using in the scanning images have the features with bimodal, through the probability of the grayscale and the Otsu's method, to gain the best threshold, and then use the threshold value to distinguish the foreground images and the background shadows, in order to realize the removing for the shadows of the scanning images. It has already achieved good effect and also lots of experimental results have demonstrated the effectiveness of the proposed method.(2) A edge-preserving image denoising methodPulse noise is the common types of image noise. This paper studied the method to remove the pulse noise of the images, and firstly use the convolution kernel in four directions to calculate the good pixels and noise pixels, then retain the information of the good pixels. As for the noise pixels in the 3×3 adjacent domain, find out the minimum distance of the good pixels, if nothing but the good pixels, the adjacent domain will expand for 5×5, then needs to search the good pixels; if the good pixels exists, replace the mid-value with the noise pixels, otherwise replace the farthest distance pixel values in adjacent domain with noise pixels. The experimental results shows that the algorithm can not only remove the image noise effectively, but also can better protect the edge of the image.(3) Edge detection using image segmentation and Sobel operatorsThe image edge detection has the value for widespread practical application, this paper researched on the multi-peak image edge extraction algorithm, and firstly, through all the grayscale expectations to confirm the area of the image segmentation thresholds, Using the maximum variance (Otsu) method combined with the expectation to calculate the image segmentation threshold, using the morphological method to eliminate isolated points and small cavities for the image that has been segmented. Finally using Sobel operators to undergo the edge detection respectively for the segmented image and the original image, adding the two edge images can be obtained the result of the edge detection. Experiments show that this algorithm less of missing inspection and accurate positioning the image edge.(4)The image edge detection based on the morphology and enhancement algorithmThe morphology is a simply and effective method in the area of image processing. Due to poor effect on the weak edge of the edge images detect by morphology, this paper detected that the edge images has characteristics of single-peak which based on the morphologic, and find out the minimum and maximum values in the histograms, through these two points available a straight line, then find out all points to a maximum distance of this line, using the largest point corresponding the grayscale as the segmentation thresholds; and then use binary threshold to achieve the purpose of enhanced the contrast between the poor effect on the edge images and the edge images. Lots of experiments show that this algorithm can solve the poor effects on the weak edge, so the method of enhancing the edge image can accurately reflects the original image edge information.